3 ms·

Last year, I interviewed Ian Clarke, who started arguably the first decentralized file sharing system: freenet! https://www.youtube.com/watch?v=JWrRqUkJpMQ htt

by EGreg 4y ago

Last year, I interviewed Ian Clarke, who started arguably the first decentralized file sharing system: freenet! https://www.youtube.com/watch?v=JWrRqUkJpMQ https://www.youtube.com/watch?v=JWrRqUkJpMQ The interview -- like most of my interviews -- goes into depth about the entire space, and also touches on Locutus, his new project. PS: I like to interview people on topics related to what we build at Qbix and Intercoin. Freedom of Speech and Capitalism (with Noam Chomsky) https://www.youtube.com/watch?v=gv5mI6ClPGc https://www.youtube.com/watch?v=gv5mI6ClPGc Identity and Civic Engagement (with David Boulet) https://www.youtube.com/watch?v=GRzVj2W9WGM https://www.youtube.com/watch?v=GRzVj2W9WGM Free Cities (Patri Friedman, grandson of Milton Friedman) https://www.youtube.com/watch?v=Lgil1M9tAXU https://www.youtube.com/watch?v=Lgil1M9tAXU Community Economics (Thomas Greco) https://www.youtube.com/watch?v=OXTn52kL0Yo https://www.youtube.com/watch?v=OXTn52kL0Yo Crypto and Securities Laws (Sara Hanks) https://www.youtube.com/watch?v=ocrqgkJn4m0 https://www.youtube.com/watch?v=ocrqgkJn4m0

Diederich 4y ago
Really cool stuff, thank you for this!
ArneBab 4y ago
Freenet has a different concept than MaidSAFE: While MaidSAFE (as it was when I last looked into it — a few years ago) is a hosting and application platform, the focus of Freenet is to be the technical component for censorship resistance. Everything it provides comes from this — even the recently added video and Android support (because Black Lives Matter showed clearly that the information that people need to share the most is video content nowadays). Therefore the overarching design goal of freenet is to make censorship as hard as technically possible. That’s the reason for - providing anonymity (else you could be threatened with repercussions - as seen in the case of the wikileaks informer from the army in the USA), - building it as free software network (else you could just shut down the central website, as people tried with wikileaks), - providing safe pseudonyms and caching of the content on all participating nodes (else people could censor by spamming or overloading nodes), - including good moderation tools (otherwise people could censor by just "flooding the zone with shit"), and even - the darknet mode and enhancements in usability (else freenet could be stopped by just prosecuting everyone who uses it, or it would reach too few people to be able to counter censorship in the clearnet). One core difference between MaidSAFE is that Freenet provides communication. Content in Freenet stays accessible while people access it, without tracking accesses: When you upload something, you can imagine that a clock starts ticking. After a while (depending on file size) the content is no longer accessible. But if someone accesses it, their node "heals" that content, which resets that clock. Also Freenet enables people to connect purely to their friends, so there’s no central point of failure or detection. I once did a comparison to other projects for a funding proposal (that did not succeed, though) which should still give you an idea about the difference between Freenet and other systems: http://www.draketo.de/proj/freenet-funding/#sec-9 http://www.draketo.de/proj/freenet-funding/#sec-9
notacoward 4y ago
Having been in that space at the time, I think this point is not arguable at all. Freenet was not the first decentralized file sharing system. Freenet might have been the subject of a PhD thesis in 1999, but by the time it became even slightly usable Napster and Gnutella already had substantial user bases. (Similar to Ceph BTW which had a long gap between being a PhD thesis and being an actual usable distributed filesystem, during which time others had existed in the wild for some time.) The "arguable" part is whether you could include eDonkey, GNUnet, and others in that list. What is true AFAIK is that Freenet was the first such system with things like censorship resistance as a primary design goal (though the MojoNation folks might even disagree with that) and a non-trivial user base. In case anyone doesn't believe me, here's the Freenet thesis which actually mentions both Gnutella and Napster in references. Hard to claim precedence when that's the case. https://www.cs.cornell.edu/people/egs/615/freenet.pdf https://www.cs.cornell.edu/people/egs/615/freenet.pdf
- s4n1ty 4y ago
  [dead]
- sanity31415 4y ago
  Napster wasn't decentralized, it relied on a centralized index which is why it was shut down. Gnutella relied on a broadcast search which didn't scale. Freenet's use of small world routing to be both decentralized and scalable was unique, as was it's use of cryptographic contracts (signed subspace keys). That paper isn't the original Freenet paper, it came later. This is the original: https://www.semanticscholar.org/paper/A-Distributed-Decentralised-Information-Storage-and-Clarke/f3ec2adb51b823e4ba7f8c86315c17c5547f88cc https://www.semanticscholar.org/paper/A-Distributed-Decentra...
  - ArneBab 4y ago
    Gnutella moved from broadcast to dynamically structured search early on (in version 0.6, 2002). See https://rfc-gnutella.sourceforge.net/src/rfc-0_6-draft.html https://rfc-gnutella.sourceforge.net/src/rfc-0_6-draft.html Including decentralized download via HTTP using just three added headers and Tiger Tree Hash Exchange: https://rfc-gnutella.sourceforge.net/src/Partial_File_Sharing_Protocol_1.0.txt https://rfc-gnutella.sourceforge.net/src/Partial_File_Sharin... Scalability was limited to global search in around 5 million nodes (with the network spanning 40 million people at its peak), though, due to restrictions in the number of TCP connections Windows (private edition) supported.
heavyset_go 4y ago
Great content, by the way.

by EGreg 4y ago

Diederich 4y ago

Really cool stuff, thank you for this!

ArneBab 4y ago

Freenet has a different concept than MaidSAFE: While MaidSAFE (as it was when I last looked into it — a few years ago) is a hosting and application platform, the focus of Freenet is to be the technical component for censorship resistance. Everything it provides comes from this — even the recently added video and Android support (because Black Lives Matter showed clearly that the information that people need to share the most is video content nowadays). Therefore the overarching design goal of freenet is to make censorship as hard as technically possible. That’s the reason for - providing anonymity (else you could be threatened with repercussions - as seen in the case of the wikileaks informer from the army in the USA), - building it as free software network (else you could just shut down the central website, as people tried with wikileaks), - providing safe pseudonyms and caching of the content on all participating nodes (else people could censor by spamming or overloading nodes), - including good moderation tools (otherwise people could censor by just "flooding the zone with shit"), and even - the darknet mode and enhancements in usability (else freenet could be stopped by just prosecuting everyone who uses it, or it would reach too few people to be able to counter censorship in the clearnet). One core difference between MaidSAFE is that Freenet provides communication. Content in Freenet stays accessible while people access it, without tracking accesses: When you upload something, you can imagine that a clock starts ticking. After a while (depending on file size) the content is no longer accessible. But if someone accesses it, their node "heals" that content, which resets that clock. Also Freenet enables people to connect purely to their friends, so there’s no central point of failure or detection. I once did a comparison to other projects for a funding proposal (that did not succeed, though) which should still give you an idea about the difference between Freenet and other systems: http://www.draketo.de/proj/freenet-funding/#sec-9 http://www.draketo.de/proj/freenet-funding/#sec-9

notacoward 4y ago

Having been in that space at the time, I think this point is not arguable at all. Freenet was not the first decentralized file sharing system. Freenet might have been the subject of a PhD thesis in 1999, but by the time it became even slightly usable Napster and Gnutella already had substantial user bases. (Similar to Ceph BTW which had a long gap between being a PhD thesis and being an actual usable distributed filesystem, during which time others had existed in the wild for some time.) The "arguable" part is whether you could include eDonkey, GNUnet, and others in that list. What is true AFAIK is that Freenet was the first such system with things like censorship resistance as a primary design goal (though the MojoNation folks might even disagree with that) and a non-trivial user base. In case anyone doesn't believe me, here's the Freenet thesis which actually mentions both Gnutella and Napster in references. Hard to claim precedence when that's the case. https://www.cs.cornell.edu/people/egs/615/freenet.pdf https://www.cs.cornell.edu/people/egs/615/freenet.pdf

s4n1ty 4y ago
[dead]
sanity31415 4y ago
Napster wasn't decentralized, it relied on a centralized index which is why it was shut down. Gnutella relied on a broadcast search which didn't scale. Freenet's use of small world routing to be both decentralized and scalable was unique, as was it's use of cryptographic contracts (signed subspace keys). That paper isn't the original Freenet paper, it came later. This is the original: https://www.semanticscholar.org/paper/A-Distributed-Decentralised-Information-Storage-and-Clarke/f3ec2adb51b823e4ba7f8c86315c17c5547f88cc https://www.semanticscholar.org/paper/A-Distributed-Decentra...
- ArneBab 4y ago
  Gnutella moved from broadcast to dynamically structured search early on (in version 0.6, 2002). See https://rfc-gnutella.sourceforge.net/src/rfc-0_6-draft.html https://rfc-gnutella.sourceforge.net/src/rfc-0_6-draft.html Including decentralized download via HTTP using just three added headers and Tiger Tree Hash Exchange: https://rfc-gnutella.sourceforge.net/src/Partial_File_Sharing_Protocol_1.0.txt https://rfc-gnutella.sourceforge.net/src/Partial_File_Sharin... Scalability was limited to global search in around 5 million nodes (with the network spanning 40 million people at its peak), though, due to restrictions in the number of TCP connections Windows (private edition) supported.

heavyset_go 4y ago

Great content, by the way.